Officially Reviewed by the Government of Luxembourg True life-saving AI isn’t the fastest — it’s the one that never stops. It isn’t the most complex, but the most ethical, transparent, and repairable that truly serves humanity. Only AI that runs on simple solar power and old smartphones can reach those who need it most — and keep them alive.
Space or left click to take over. The AI stops acting until the next round.I am Gyumin Jeon, an independent researcher from South Korea with a background in national universities, dedicated to non-profit and public-interest research.
EU Legal & Human Rights Vetting (Luxembourg Official Review)
The project’s legal and ethical framework was formally reviewed by a G7/EU government. The Government of Luxembourg—through its Ministry of Foreign Affairs and LuxDev (the Luxembourg Development Cooperation Agency)—conducted a three-week review confirming that this system fully aligns with the EU’s rigorous legal standards (GDPR) and core human rights principles.
The review concluded that the project did not fall within Luxembourg’s Development Cooperation priorities. However, this outcome further clarified the project’s true purpose: not merely as traditional aid, but as a framework for 'Technical Sovereignty' that empowers local communities to own and operate their own infrastructure. This makes it far better suited for in‑depth legal, policy, and ethical research partnerships with law schools and advanced academic institutions worldwide.
In today’s AI landscape, competition often centers on larger and more complex models. Yet what humanity truly needs is an intelligence that is “small enough, yet smart enough.” This system embodies that philosophy, combining five independent AI engines that together remain under 100 kilobytes—a feat of minimalism and resilience.
In the first stage, the 54 KB Offline AI collects and normalizes TinyML data, handling foundational inputs and environmental signals. This process operates completely offline, free from any dependence on network connectivity or cloud infrastructure.
In the second stage, the Dual-Brain Micro-AI interprets and integrates the primary data to perform higher-level decision-making and adaptive learning. This dual-layer structure is implemented hierarchically at the code level, mirroring the human brain’s division between analytical and intuitive reasoning.
As a result, the system operates reliably even in low-power solar environments or on older smartphones, demonstrating remarkably high computational efficiency and long-term stability. More than a technical experiment, it serves as proof that true AI autonomy is achievable in developing regions, refugee camps, and areas with limited electricity access.
This architecture represents not “bigger AI,” but “more sustainable AI.” Built on ethics, accessibility, and practical resilience, this ultra-lightweight intelligence stands as a foundation for a new era of humanitarian diplomacy and inclusive technological development.
An ethical offline AI overview PDF aligned with EU standards — an EU-vetted ethical framework for non-commercial AI, explaining how an ultra-lightweight offline ethical AI can be applied in the field, as presented to the Government of Luxembourg. The Mouse/Refugee games demonstrate ultra-lightweight, adaptive AI designed to operate on solar-powered, off-grid devices. The Malaria/Ticker demos showcase “Numerical AI,” which uses only numbers and labels—no text—to minimize legal risks for NGOs. As explained in the Saving Lives with Ethical AI PDF, this method is used to accurately tally field resource needs such as food and medicine. (This represents only a very small portion of the communications exchanged with the Government of Luxembourg.)